Federated AI for Autonomous Vehicles: SSL & Distillation

RISE Research Institutes of Sweden; Kista; On-site; SEK 35,000 - 43,000; Part time

Benefits offered by this job

On-site at RISE Kista

Job summary

RISE Research Institutes of Sweden seeks two highly motivated Master’s students to contribute to DREAM, researching federated self-supervised learning and knowledge distillation for autonomous vehicles. You will work closely with the RISE and Zenseact teams, and be expected to be on-site at the Kista office at least three days per week.

The project focuses on real-time federated learning approaches, exploring self-supervised methods and cross-model knowledge transfer, with opportunities to work

Qualifications

• Strong background in machine learning and computer vision.
• Experience with federated learning, self-supervised learning, knowledge distillation or autonomous-driving data is a merit.
• On-site work at RISE Kista office at least three days per week.

Responsibilities

• Work on Topic 1 – Federated SSL: develop and evaluate novel SSL methods for driving data and compare with baselines.
• Work on Topic 2 – Federated Knowledge Transfer: study distillation/transfer across heterogeneous models and adapt to changes in architecture/sensors/hardware.
• Implement and evaluate proposed methods using autonomous-driving data (ZOD) and present results to DREAM partners.

Skills

Deep learning

Python programming

Literature review

Federated learning

Self-supervised learning

Knowledge distillation

Autonomous-driving data

Education

Master's student in ML/CV or related field

Job description

RISE Research Institutes of Sweden seeks two highly motivated Master’s students to contribute to DREAM, researching federated self-supervised learning and knowledge distillation for autonomous vehicles. You will work closely with the RISE and Zenseact teams, and be expected to be on-site at the Kista office at least three days per week.

The project focuses on real-time federated learning approaches, exploring self-supervised methods and cross-model knowledge transfer, with opportunities to work